i-tag-u | u-tag-me
may 04, 2008.
I allow 'social music' revolutionaries LastFM
to track my listening habits, not only by using their streaming
music facilities, but also by enabling the audioscrobbler
to chart most of the music that I play on my laptop.
(Most, and not all of it, because the 'scrobbler' can only handle
data from my iTunes player, and not that from the other software audio players
that I use intermittently. Moreover, the 'scrobbler' can not handle audio
streams, also not when I play them in iTunes. But almost always
when I listen to an iTune, audioscrobbler faithfully sends the
track's metadata to the LastFM database, and files them there as part of
my user's profile. As I also play music regularly through other than software
means (CD-, record-, cassette-players), the gathered data of course only
can give a partial image, but still, there's a lot of info going 'their
way' day by day ...)
As a result LastFM's database at the time of this writing includes a list with the 'metadata' that
are part of
some 10.000
music files that
I have listened to over the past two years, while they were 'watching' ...
(It contains similar information pertaining to its - as claimed by
the site - over 19 million other regular users. 19 million ... so, no, indeed,
"just turn on with me and you're not alone ...", as one of our favorite
glam-rockers used to sing ... It could've been LastFM's motto (
* ) ...)
Here's what my 'musical profile' at LastFM's looks like when you present it as a tagcloud. It was extracted from my listening data as they have been stocked over the past two years by LastFM :
The tags in this 'cloud' (ordered alphabetically as you see) are words and phrases used by LastFM listeners to classify the music of the fifty artists that I listened to most over the past two years. Of course some words will be used more than others, and LastFM assigns 'weights' to them, which correspond to the number of times that the tag was used to 'describe' the artist. One may then sum the relative weights of a given artist-tag over the tag's occurrences in my artist top 50, where each occurrence in turn is weighted by the number of tracks of the artist that I played. This constitutes my tag-vector, in which each dimension is a tag, and each dimension's value the relative importance of the tag within my top 50: a given dimension has a large value if it corresponds to a tag that is commonly used by LastFM users to classify the music of many of the artists that I listened to most.
Well, you see above how this little astuce worked out ... What do you think? Is it me? Do you or don't you recognize me there? Whaddya say?
It is possible to do this sort of 'number crunching' yourself, because the Audioscrobbler site gives you, under a creative commons 'non-commercial' license, access to quite a bit of LastFM's database. It was not me, though, who painted this succinct portrait of Har$' musical tastes and preferences. It was produced by a sweet little online script written by Anthony Liekens (from Borsbeek in Belgium). For those of you that also have a LastFM account, when you follow the link to his website, your own 'musical tag cloud' is but a click away ...
If you click on the tags in the cloud, you see that all of them lead to soundblog entries or similar pages. Now of course these links were added by myself. I've come to appreciate a lot the 'power' of tagclouds as a navigational tool, so indeed I jumped at the possibility to use this particular cloud and its tags to send you way back into corners, crypts and caverns of this blog, for which over the five years of its existence (indeed, already pushing six by now ...) I have scribbled some three hundred thousand words, most of which most viewers will never see, let alone read ... So while she's here, why not have her click ? ... on ... and have her click ?... on ... clicky-click-clack-cluck-click ... on ... Also Liekens' script generates links, but these lead to LastFM 'tag-pages'. I will let you investigate these pages by yourself. They provide a list of the artists that have been most 'tagged' with a particular word or expression. On the page for the tag 'experimental' one finds for example that - at the time of this writing - Björk is the number one artist thus tagged at LastFM, while Philip Glass ranks topmost among contemporary classical artists, a list that also includes Igor Stravinsky. All of this is not so strange, given that obviously a very large majority of LastFM listeners will listen mostly to pop music that does not deviate too much from some main stream.
About a year ago Anthony Liekens used the public Audioscrobbler base to do some data-mining. He analyzed a sample group of 2480 LastFM users (picked, I guess, randomly), and found that roughly this sample could be partitioned into five 'musical preference' groups, with little or no overlap: electronic/pop, rock, indie, metal, hip-hop. That is sort of surprising, especially the observation that these five groups appear to be near to mutually 'exclusive'. It also hints once again at the potential of 'mining' in such 'data dumps' ... It's a treasure stove, that all users of this 'webservice' have been building and continue to build together for LastFM ... Not in the least from a marketing technical point of view, as LastFM indeed is able to reach, in one single scripted instruction, all and only its users that recently have been listening to a certain type of music, a specific band, or even a certain track ... and do such things within certain geographical regions ... et cetera.
Like for most of the 'tagging' as it is being practiced by users in the context of the global 'social web' in order to classify and access a mindboggling quantity of information and media, there are no fixed rules for this keyword-ing of music and musicians. There is no fixed list of tags. Anyone may use any word or whatever sequence of symbols as a label. So there is avantgarde as well as avant-garde, people tag artists as seen live, or as crap; as silver or gold; as web or as spider ... The fact that nevertheless this way of freely labeling and classifying content turns out to be so very effective as a means to access large quantities of media-content (sound, images, text ...) may at first sight seem somewhat surprising. But of course it is not, as soon as there are a great number of taggers involved. Large numbers filter out the individual idiosyncrasies, and make 'social tagging' work because we all speak the 'same' language, and share a certain 'semantics'. (It actually is fun to play from time to time a LastFM station, exclusively filled with music tagged crap, or shit, or sucks ... or other curious tags. Not for too long, though, but there's surprises, off and on, here and there ...)
As several others, I found LastFM's approach to music recommendation, based as it is upon a notion of 'similar taste' - "if both Jeff and Lisa like the Ants a lot, and Jeff also plays a lot of Reverend Tiger, then also Lisa is likely to appreciate the Reverend Tiger" - to be far more efficient in 'proposing' me music that I liked but did not know yet, than the one based upon a classification of tracks according to a long list of more or less formal musical parameters (attributes or genes that come together in the socalled 'Musical Genome'), as used by Pandora.com. At the time that I listened to Pandora's 'recommendations' I did quite a bit of skipping and banning of tracks, which I seem to be doing only rarely when listening to LastFM's 'similar artists' stations. Since Pandora (for reasons of licensing constraints) had to stop delivering streaming audio outside of the US already over a year ago, it is no longer possible to keep testing the relative merits of both.
With its catalogue of streamable music in any imaginable genre getting ever more complete, over the past couple of years LastFM rapidly became more interesting and more and more useful, and maybe the same holds true for Pandora. It is a shame that when based in Europe at the moment you can only use its services if you manage to hide your (european) IP address in one way or another.
It'll be interesting to see, and I actually wonder, whether LastFM will be able to continue to expand this world's largest collection of online music, and serve it for free ( ** ) in this 'self regulating' way, principally based upon the evolving 'taste profiles' of its users. Or will LastFM become evermore of a 'music pusher', instead of a 'music recommender'? It is possible already for artists and record companies to pay LastFM to 'recommend' certain tracks to groups of users with a specific profile. That's, quite appropriately, called 'PowerPlay'. According to the information on the LastFM site in its 'Music Manager' sections, it'll cost you (or rather: it would cost me (;- ...) €165 to have a certain track played a thousand times as part of the streams served to a specific group of users. Two things here that bother me: first that this of course interferes with the dynamics of the system of recommendations based upon similarities of profiles and listening habits, a system that I like and the outcomes of which I appreciate; and second that there is no way (yet?) to know which of the musics I am listening too I am hearing because of this (payment that therefore is like) payola ... That's bad; formerly when listening to traditional (First)FM radio, at least one knew more or less on what stations plays were likely to have been bought and by who ... LastFM should start indicating which plays continue to turn up in a stream because someone paid for them to do so. If not, it will (and it should) quickly loose its credibility. I actually do suspect that paid-for plays of tracks are turning up in the recommendation streams generated also for margin profiles like mine ...
...
All through school there was this varying but always kind of big group of kids that spent endless time keeping track of and comparing 'the top of the pops', by following, and producing, 'hit' charts in all sorts of variations. They must be all on LastFM now, which is the ultimate chart-machine. The sheer endless opportunities LastFM offers for statistic (self-)indulgence through ordered lists of all sorts, that one may consult and generate there, is one of the main reasons for its evergrowing popularity ... They are well aware of that ...
For, you see ... my eclectic score is a proud 96/100 ! ...
notes __ ::
(*) It's a line from David Bowie's Rock 'n' Roll
Suicide, a track from his 1972 The
Rise and Fall of Ziggy Stardust and the Spiders from Mars album ...
[ ^ ]
(**) LastFM actually does have a subscription
service, which, for two and a half euros monthly offers a number of advantages,
including ad-free surfing of the web-site. I doubt that ad-free includes
'free from paid-for promotion plays'. The site actually also since may 2007
no longer is independent, but it is owned by CBS
Interactive, an acquisition that however - at least as far as I am concerned
- went completely unnoticed as to the quality of my LastFM 'experience'.
If anything, I'd say that experience on the contrary improved ... [
^ ]
[ Earlier related SB-entry: reci[-di?-]va ]
tags: tagcloud, internet radio, LastFM
# .260.
comments for « i-tag-u | u-tag-me » ::
|
Comments are disabled |
